Due to the rapid growth in technology employed by the spammers, there is a need of classifiers that are more efficient, generic and highly adaptive. Neural Network based technologies have high ability of adaption as well as generalization. As per our knowledge, very little work has been done in this field using neural network. We present this paper to fill this gap. This paper evaluates performance of three supervised learning algorithms of artificial neural network by creating classifiers for the complex problem of latest web spam pattern classification. These algorithms are Conjugate Gradient algorithm, Resilient Backpropagation learning, and Levenberg-Marquardt algorithm.
ARTIFICIAL NEURAL NETWORK APPROACH TO MODELING OF POLYPROPYLENE REACTORijac123
This paper shows modeling of highly nonlinear polymerization process using the artificial neural network approach for the model predictive purposes. Polymerization occurs in a fluidized bed polypropylene reactor using Ziegler - Natta catalyst and the main objective was modeling of the reactor production rate.
The data set used for an identification of the model is a real process data received from an existing polypropylene plant and the identified model is a nonlinear autoregressive neural network with the exogenous input. Performance of a trained network has been verified using the real process data and the
ability of the production rate prediction is shown in the conclusion.
Implementation Of Back-Propagation Neural Network For Isolated Bangla Speech ...ijistjournal
This paper is concerned with the development of Back-propagation Neural Network for Bangla Speech Recognition. In this paper, ten bangla digits were recorded from ten speakers and have been recognized.The features of these speech digits were extracted by the method of Mel Frequency Cepstral Coefficient(MFCC) analysis. The mfcc features of five speakers were used to train the network with Back propagation algorithm. The mfcc features of ten bangla digit speeches, from 0 to 9, of another five speakers were used to test the system. All the methods and algorithms used in this research were implemented using the features of Turbo C and
C++ languages. From our investigation it is seen that the developed system can successfully encode and analyze the mfcc features of the speech signal to recognition. The developed system achieved recognition rate about 96.332% for known speakers (i.e., speaker dependent) and 92% for unknown speakers (i.e., speaker independent).
Implementation of Back-Propagation Neural Network using Scilab and its Conver...IJEEE
Artificial neural network has been widely used for solving non-linear complex tasks. With the development of computer technology, machine learning techniques are becoming good choice. The selection of the machine learning technique depends upon the viability for particular application. Most of the non-linear problems have been solved using back propagation based neural network. The training time of neural network is directly affected by convergence speed. Several efforts are done to improve the convergence speed of back propagation algorithm. This paper focuses on the implementation of back-propagation algorithm and an effort to improve its convergence speed. The algorithm is written in SCILAB. UCI standard data set is used for analysis purposes. Proposed modification in standard backpropagation algorithm provides substantial improvement in the convergence speed.
ARTIFICIAL NEURAL NETWORK APPROACH TO MODELING OF POLYPROPYLENE REACTORijac123
This paper shows modeling of highly nonlinear polymerization process using the artificial neural network approach for the model predictive purposes. Polymerization occurs in a fluidized bed polypropylene reactor using Ziegler - Natta catalyst and the main objective was modeling of the reactor production rate.
The data set used for an identification of the model is a real process data received from an existing polypropylene plant and the identified model is a nonlinear autoregressive neural network with the exogenous input. Performance of a trained network has been verified using the real process data and the
ability of the production rate prediction is shown in the conclusion.
Implementation Of Back-Propagation Neural Network For Isolated Bangla Speech ...ijistjournal
This paper is concerned with the development of Back-propagation Neural Network for Bangla Speech Recognition. In this paper, ten bangla digits were recorded from ten speakers and have been recognized.The features of these speech digits were extracted by the method of Mel Frequency Cepstral Coefficient(MFCC) analysis. The mfcc features of five speakers were used to train the network with Back propagation algorithm. The mfcc features of ten bangla digit speeches, from 0 to 9, of another five speakers were used to test the system. All the methods and algorithms used in this research were implemented using the features of Turbo C and
C++ languages. From our investigation it is seen that the developed system can successfully encode and analyze the mfcc features of the speech signal to recognition. The developed system achieved recognition rate about 96.332% for known speakers (i.e., speaker dependent) and 92% for unknown speakers (i.e., speaker independent).
Implementation of Back-Propagation Neural Network using Scilab and its Conver...IJEEE
Artificial neural network has been widely used for solving non-linear complex tasks. With the development of computer technology, machine learning techniques are becoming good choice. The selection of the machine learning technique depends upon the viability for particular application. Most of the non-linear problems have been solved using back propagation based neural network. The training time of neural network is directly affected by convergence speed. Several efforts are done to improve the convergence speed of back propagation algorithm. This paper focuses on the implementation of back-propagation algorithm and an effort to improve its convergence speed. The algorithm is written in SCILAB. UCI standard data set is used for analysis purposes. Proposed modification in standard backpropagation algorithm provides substantial improvement in the convergence speed.
Fundamental, An Introduction to Neural NetworksNelson Piedra
An introduction to Neural Networks, eight edition, 1996
Authors: Ben Krose, Faculty of Mathematics & Computer Science, University of Amsterdam. Patrick wan der Smagt, Institute of Robotics and Systems Dynamics, German Aerospace Research Establishment
Keynote: Nelson Piedra, Computer Sciences School - Advanced Tech, Technical University of Loja UTPL, Ecuador.
Here is my class on the multilayer perceptron where I look at the following:
1.- The entire backproagation algorithm based in the gradient descent
However, I am planning the tanning based in Kalman filters.
2.- The use of matrix computations to simplify the implementations.
I hope you enjoy it.
Artificial neural network for machine learninggrinu
An Artificial Neurol Network (ANN) is a computational model. It is based on the structure and functions of biological neural networks. It works like the way human brain processes information. ANN includes a large number of connected processing units that work together to process information. They also generate meaningful results from it.
Deep Learning and Tensorflow Implementation(딥러닝, 텐서플로우, 파이썬, CNN)_Myungyon Ki...Myungyon Kim
Deep learning and Tensorflow implementation
2016.11.16
<Cotents>
Feature Engineering
Deep Neural Network
Tensorflow
Tensorflow Implementation
Future works
References
This slides deals with several things about deep learning.
ex) History of Deep learning, Several difficulties and breakthroughs. Things related to deep learning such as activation functions, perceptrons, Backpropagation, pre-train, drop-out, Convolutional Neural Network (CNN), Simple implementation of Tensor Flow, Python, and so on.
딥러닝, 기계학습, 머신러닝, 텐서플로우, 파이썬
LOCALIZATION OF OVERLAID TEXT BASED ON NOISE INCONSISTENCIESaciijournal
In this paper, we present a novel technique for localization of caption text in video frames based on noise inconsistencies. Text is artificially added to the video after it has been captured and as such does not form part of the original video graphics. Typically, the amount of noise level is uniform across the entire captured video frame, thus, artificially embedding or overlaying text on the video introduces yet another segment of noise level. Therefore detection of various noise levels in the video frame may signify availability of overlaid text. Hence we exploited this property by detecting regions with various noise levels to localize overlaid text in video frames. Experimental results obtained shows a great improvement in line with overlaid text localization, where we have performed metric measure based on Recall, Precision and f-measure.
TRUST METRICS IN RECOMMENDER SYSTEMS: A SURVEYaciijournal
Information overload is a new challenge in e-commerce sites. The problem refers to the fast growing of
information that lead following the information flow in real world be impossible. Recommender systems, as
the most successful application of information filtering, help users to find items of their interest from huge
datasets. Collaborative filtering, as the most successful technique for recommendation, utilises social
behaviours of users to detect their interests. Traditional challenges of Collaborative filtering, such as cold
start, sparcity problem, accuracy and malicious attacks, derived researchers to use new metadata to
improve accuracy of recommenders and solve the traditional problems. Trust based recommender systems
focus on trustworthy value on relation among users to make more reliable and accurate recommends. In
this paper our focus is on trust based approach and discuss about the process of making recommendation
in these method. Furthermore, we review different proposed trust metrics, as the most important step in this
process.
Fundamental, An Introduction to Neural NetworksNelson Piedra
An introduction to Neural Networks, eight edition, 1996
Authors: Ben Krose, Faculty of Mathematics & Computer Science, University of Amsterdam. Patrick wan der Smagt, Institute of Robotics and Systems Dynamics, German Aerospace Research Establishment
Keynote: Nelson Piedra, Computer Sciences School - Advanced Tech, Technical University of Loja UTPL, Ecuador.
Here is my class on the multilayer perceptron where I look at the following:
1.- The entire backproagation algorithm based in the gradient descent
However, I am planning the tanning based in Kalman filters.
2.- The use of matrix computations to simplify the implementations.
I hope you enjoy it.
Artificial neural network for machine learninggrinu
An Artificial Neurol Network (ANN) is a computational model. It is based on the structure and functions of biological neural networks. It works like the way human brain processes information. ANN includes a large number of connected processing units that work together to process information. They also generate meaningful results from it.
Deep Learning and Tensorflow Implementation(딥러닝, 텐서플로우, 파이썬, CNN)_Myungyon Ki...Myungyon Kim
Deep learning and Tensorflow implementation
2016.11.16
<Cotents>
Feature Engineering
Deep Neural Network
Tensorflow
Tensorflow Implementation
Future works
References
This slides deals with several things about deep learning.
ex) History of Deep learning, Several difficulties and breakthroughs. Things related to deep learning such as activation functions, perceptrons, Backpropagation, pre-train, drop-out, Convolutional Neural Network (CNN), Simple implementation of Tensor Flow, Python, and so on.
딥러닝, 기계학습, 머신러닝, 텐서플로우, 파이썬
LOCALIZATION OF OVERLAID TEXT BASED ON NOISE INCONSISTENCIESaciijournal
In this paper, we present a novel technique for localization of caption text in video frames based on noise inconsistencies. Text is artificially added to the video after it has been captured and as such does not form part of the original video graphics. Typically, the amount of noise level is uniform across the entire captured video frame, thus, artificially embedding or overlaying text on the video introduces yet another segment of noise level. Therefore detection of various noise levels in the video frame may signify availability of overlaid text. Hence we exploited this property by detecting regions with various noise levels to localize overlaid text in video frames. Experimental results obtained shows a great improvement in line with overlaid text localization, where we have performed metric measure based on Recall, Precision and f-measure.
TRUST METRICS IN RECOMMENDER SYSTEMS: A SURVEYaciijournal
Information overload is a new challenge in e-commerce sites. The problem refers to the fast growing of
information that lead following the information flow in real world be impossible. Recommender systems, as
the most successful application of information filtering, help users to find items of their interest from huge
datasets. Collaborative filtering, as the most successful technique for recommendation, utilises social
behaviours of users to detect their interests. Traditional challenges of Collaborative filtering, such as cold
start, sparcity problem, accuracy and malicious attacks, derived researchers to use new metadata to
improve accuracy of recommenders and solve the traditional problems. Trust based recommender systems
focus on trustworthy value on relation among users to make more reliable and accurate recommends. In
this paper our focus is on trust based approach and discuss about the process of making recommendation
in these method. Furthermore, we review different proposed trust metrics, as the most important step in this
process.
Audio Signal Identification and Search Approach for Minimizing the Search Tim...aciijournal
Audio or music fingerprints can be utilize to implement an economical music identification system on a
million-song library, however the system needs great deal of memory to carry the fingerprints and indexes.
Therefore, for a large-scale audio library, memory
imposes a restriction on the speed of music
identifications. So, we propose an efficient music
identification system which used a kind of space-saving
audio fingerprints. For saving space, original finger representations are sub-sample and only one quarters
of the original data is reserved. In this approach,
memory demand is far reduced and therefore the search
speed is criticalincreasing whereas the lustiness and dependability are well preserved. Mapping
audio information to time and frequency domain for
the classification, retrieval or identification tasks
presents four principal challenges. The dimension o
f the input should be considerably reduced;
the ensuing options should be strong to possible distortions of the input; the feature should be informative
for the task at hand simple. We propose distortion
free system which fulfils all four of these requirements.
Extensive study has been done to compare our system
with the already existing ones, and the results sh
ow
that our system requires less memory, provides fast
results and achieves comparable accuracy for a large-
scale database.
Text mining is the process of extracting interesting and non-trivial knowledge or information from unstructured text data. Text mining is the multidisciplinary field which draws on data mining, machine learning, information retrieval, computational linguistics and statistics. Important text mining processes are information extraction, information retrieval, natural language processing, text classification, content analysis and text clustering. All these processes are required to complete the preprocessing step before doing their intended task. Pre-processing significantly reduces the size of the input text documents and the actions involved in this step are sentence boundary determination, natural language specific stop-word elimination, tokenization and stemming. Among this, the most essential and important action is the tokenization. Tokenization helps to divide the textual information into individual words. For performing tokenization process, there are many open source tools are available. The main objective of this work is to analyze the performance of the seven open source tokenization tools. For this comparative analysis, we have taken Nlpdotnet Tokenizer, Mila Tokenizer, NLTK Word Tokenize, TextBlob Word Tokenize, MBSP Word Tokenize, Pattern Word Tokenize and Word Tokenization with Python NLTK. Based on the results, we observed that the Nlpdotnet Tokenizer tool performance is better than other tools.
Recommender systems have grown to be a critical research subject after the emergence of the first paper on collaborative filtering in the Nineties. Despite the fact that educational studies on recommender systems, has extended extensively over the last 10 years, there are deficiencies in the complete literature evaluation and classification of that research. Because of this, we reviewed articles on recommender structures, and then classified those based on sentiment analysis. The articles are categorized into three techniques of recommender system, i.e.; collaborative filtering (CF), content based and context based. We have tried to find out the research papers related to sentimental analysis based recommender system. To classify research done by authors in this field, we have shown different approaches of recommender system based on sentimental analysis with the help of tables. Our studies give statistics, approximately trends in recommender structures research, and gives practitioners and researchers with perception and destiny route on the recommender system using sentimental analysis. We hope that this paper enables all and sundry who is interested in recommender systems research with insight for destiny.
Check out the exclusive range of wall wallpaper we have to offer. Browse through our wide range of wallpapers and pick according to your requirements. Contact us now!
Learning strategy with groups on page based students' profilesaciijournal
Most of students desire to know about their knowledge level to perfect their exams. In learning environment the fields of study overwhelm on page with collaboration or cooperation. Students can do their exercises either individually or collaboratively with their peers. The system provides the guidelines for students' learning system about interest fields as Java in this system. Especially the system feedbacks information about exam to know their grades without teachers. The participants who answered the exam can discuss with each others because of sharing e mail and list of them.
Erca energy efficient routing and reclusteringaciijournal
The pervasive application of wireless sensor networks (WNSs) is challenged by the scarce energy constraints of sensor nodes. En-route filtering schemes, especially commutative cipher based en-route filtering (CCEF) can saves energy with better filtering capacity. However, this approach suffer from fixed paths and inefficient underlying routing designed for ad-hoc networks. Moreover, with decrease in remaining sensor nodes, the probability of network partition increases. In this paper, we propose energy-efficient routing and re-clustering algorithm (ERCA) to address these limitations. In proposed scheme with reduction in the number of sensor nodes to certain thresh-hold the cluster size and transmission range dynamically maintain cluster node-density. Performance results show that our approach demonstrate filtering-power, better energy-efficiency, and an average gain over 285% in network lifetime.
SURVEY SURFER: A WEB BASED DATA GATHERING AND ANALYSIS APPLICATIONaciijournal
The most important need for a web based survey technology is speedy performance and accurate results
.Though there are variety of ways a survey can be taken manually and have it assessed manually but here
we are introducing the survey site concept where the assessment of the survey responses are created
automatically by applying certain statistics. In this paper we propose the idea of automated approach to
the analysis of the survey where the results would be evident graphically to take a wise decision. Additional
to graphical view we aim on forming a platform to view complex, tedious data in simple, understandable,
interactive and visualized form.
Automatic Unsupervised Data Classification Using Jaya Evolutionary Algorithmaciijournal
In this paper we attempt to solve an automatic clustering problem by optimizing multiple objectives such as
automatic k-determination and a set of cluster validity indices concurrently. The proposed automatic
clustering technique uses the most recent optimization algorithm Jaya as an underlying optimization
stratagem. This evolutionary technique always aims
to attain global best solution rather than a local best solution in larger datasets. The explorations and exploitations imposed on the proposed work results to
detect the number of automatic clusters, appropriate partitioning present in data sets and mere optimal
values towards CVIs frontiers. Twelve datasets of different intricacy are used to endorse the performance
of aimed algorithm. The experiments lay bare that the conjectural advantages of multi objective clustering optimized with evolutionary approaches decipher into realistic and scalable performance paybacks.
An Intelligent Method for Accelerating the Convergence of Different Versions ...aciijournal
In a wide range of applications, solving the linear
system of equations Ax = b is appeared. One of the
best
methods to solve the large sparse asymmetric linear
systems is the simplified generalized minimal resi
dual
(SGMRES(m)) method. Also, some improved versions of
SGMRES(m) exist: SGMRES-E(m, k) and
SGMRES-DR(m, k). In this paper, an intelligent heur
istic method for accelerating the convergence of th
ree
methods SGMRES(m), SGMRES-E(m, k), and SGMRES-DR(m,
k) is proposed. The numerical results
obtained from implementation of the proposed approa
ch on several University of Florida standard
matrixes confirm the efficiency of the proposed met
hod.
Blood groups matcher frame work based on three level rules in myanmaraciijournal
Today Blood donation is a global interest for world to be survival lives when people are in trouble because of natural disaster. The system provides the ability how to decide to donate the blood according to the rules for blood donation not to meet the physicians. In this system, there are three main parts to accept blood from donors when they want to donate according to the features like personal health. The application facilitates to negotiate between blood donors and patients who need to get blood seriously on page without going to Blood Banks and waiting time in queue there.
HYBRID DATA CLUSTERING APPROACH USING K-MEANS AND FLOWER POLLINATION ALGORITHMaciijournal
Data clustering is a technique for clustering set of objects into known number of groups. Several approaches are widely applied to data clustering so that objects within the clusters are similar and objects in different clusters are far away from each other. K-Means, is one of the familiar center based clustering algorithms since implementation is very easy and fast convergence. However, K-Means algorithm suffers from initialization, hence trapped in local optima. Flower Pollination Algorithm (FPA) is the global optimization technique, which avoids trapping in local optimum solution. In this paper, a novel hybrid data clustering approach using Flower Pollination Algorithm and K-Means (FPAKM) is proposed. The proposed algorithm results are compared with K-Means and FPA on eight datasets. From the experimental results, FPAKM is better than FPA and K-Means.
Prediction of lung cancer is most challenging problem due to structure of cancer cell, where most of the cells are overlapped each other. The image processing techniques are mostly used for prediction of lung cancer and also for early detection and treatment to prevent the lung cancer. To predict the lung cancer various features are extracted from the images therefore, pattern recognition based approaches are useful to predict the lung cancer. Here, a comprehensive review for the prediction of lung cancer by previous researcher using image processing techniques is presented. The summary for the prediction of lung cancer by previous researcher using image processing techniques is also presented.
Web Spam Classification Using Supervised Artificial Neural Network Algorithmsaciijournal
Due to the rapid growth in technology employed by the spammers, there is a need of classifiers that are
more efficient, generic and highly adaptive. Neural Network based technologies have high ability of
adaption as well as generalization. As per our knowledge, very little work has been done in this field using
neural network. We present this paper to fill this gap. This paper evaluates performance of three supervised
learning algorithms of artificial neural network by creating classifiers for the complex problem of latest
web spam pattern classification. These algorithms are Conjugate Gradient algorithm, Resilient Backpropagation learning, and Levenberg-Marquardt algorithm.
IJRET : International Journal of Research in Engineering and Technology is an international peer reviewed, online journal published by eSAT Publishing House for the enhancement of research in various disciplines of Engineering and Technology. The aim and scope of the journal is to provide an academic medium and an important reference for the advancement and dissemination of research results that support high-level learning, teaching and research in the fields of Engineering and Technology. We bring together Scientists, Academician, Field Engineers, Scholars and Students of related fields of Engineering and Technology
Modeling of neural image compression using gradient decent technologytheijes
The International Journal of Engineering & Science is aimed at providing a platform for researchers, engineers, scientists, or educators to publish their original research results, to exchange new ideas, to disseminate information in innovative designs, engineering experiences and technological skills. It is also the Journal's objective to promote engineering and technology education. All papers submitted to the Journal will be blind peer-reviewed. Only original articles will be published.
The papers for publication in The International Journal of Engineering& Science are selected through rigorous peer reviews to ensure originality, timeliness, relevance, and readability.
Theoretical work submitted to the Journal should be original in its motivation or modeling structure. Empirical analysis should be based on a theoretical framework and should be capable of replication. It is expected that all materials required for replication (including computer programs and data sets) should be available upon request to the authors.
The International Journal of Engineering & Science would take much care in making your article published without much delay with your kind cooperation
An Artificial Neural Network (ANN) is a computational model inspired by the structure and functioning of the human brain's neural networks. It consists of interconnected nodes, often referred to as neurons or units, organized in layers. These layers typically include an input layer, one or more hidden layers, and an output layer.
Neural network based numerical digits recognization using nnt in matlabijcses
Artificial neural networks are models inspired by human nervous system that is capable of learning. One of
the important applications of artificial neural network is character Recognition. Character Recognition
finds its application in number of areas, such as banking, security products, hospitals, in robotics also.
This paper is based on a system that recognizes a english numeral, given by the user, which is already
trained on the features of the numbers to be recognized using NNT (Neural network toolbox) .The system
has a neural network as its core, which is first trained on a database. The training of the neural network
extracts the features of the English numbers and stores in the database. The next phase of the system is to
recognize the number given by the user. The features of the number given by the user are extracted and
compared with the feature database and the recognized number is displayed.
Open CV Implementation of Object Recognition Using Artificial Neural Networksijceronline
International Journal of Computational Engineering Research (IJCER) is dedicated to protecting personal information and will make every reasonable effort to handle collected information appropriately. All information collected, as well as related requests, will be handled as carefully and efficiently as possible in accordance with IJCER standards for integrity and objectivity.
Review: “Implementation of Feedforward and Feedback Neural Network for Signal...IJERA Editor
Main focus of project is on implementation of Neural Network Architecture (NNA) with on chip learning on
Analog VLSI Technology for signal processing application. In the proposed paper the analog components like
Gilbert Cell Multiplier (GCM), Neuron Activation Function (NAF) are used to implement artificial NNA.
Analog components used comprises of multiplier, adder and tan sigmoidal function circuit using MOS transistor.
This Neural Architecture is trained using Back Propagation (BP) Algorithm in analog domain with new
techniques of weight storage. Layout design and verification of above design is carried out using VLSI Backend
Microwind 3.1 software Tool. The technology used to design layout is 32 nm CMOS Technology.
Adaptive modified backpropagation algorithm based on differential errorsIJCSEA Journal
A new efficient modified back propagation algorithm with adaptive learning rate is proposed to increase the convergence speed and to minimize the error. The method eliminates initial fixing of learning rate through trial and error and replaces by adaptive learning rate. In each iteration, adaptive learning rate for output and hidden layer are determined by calculating differential linear and nonlinear errors of output layer and hidden layer separately. In this method, each layer has different learning rate in each iteration. The performance of the proposed algorithm is verified by the simulation results.
Similar to Web spam classification using supervised artificial neural network algorithms (20)
Pushing the limits of ePRTC: 100ns holdover for 100 daysAdtran
At WSTS 2024, Alon Stern explored the topic of parametric holdover and explained how recent research findings can be implemented in real-world PNT networks to achieve 100 nanoseconds of accuracy for up to 100 days.
PHP Frameworks: I want to break free (IPC Berlin 2024)Ralf Eggert
In this presentation, we examine the challenges and limitations of relying too heavily on PHP frameworks in web development. We discuss the history of PHP and its frameworks to understand how this dependence has evolved. The focus will be on providing concrete tips and strategies to reduce reliance on these frameworks, based on real-world examples and practical considerations. The goal is to equip developers with the skills and knowledge to create more flexible and future-proof web applications. We'll explore the importance of maintaining autonomy in a rapidly changing tech landscape and how to make informed decisions in PHP development.
This talk is aimed at encouraging a more independent approach to using PHP frameworks, moving towards a more flexible and future-proof approach to PHP development.
Elevating Tactical DDD Patterns Through Object CalisthenicsDorra BARTAGUIZ
After immersing yourself in the blue book and its red counterpart, attending DDD-focused conferences, and applying tactical patterns, you're left with a crucial question: How do I ensure my design is effective? Tactical patterns within Domain-Driven Design (DDD) serve as guiding principles for creating clear and manageable domain models. However, achieving success with these patterns requires additional guidance. Interestingly, we've observed that a set of constraints initially designed for training purposes remarkably aligns with effective pattern implementation, offering a more ‘mechanical’ approach. Let's explore together how Object Calisthenics can elevate the design of your tactical DDD patterns, offering concrete help for those venturing into DDD for the first time!
Smart TV Buyer Insights Survey 2024 by 91mobiles.pdf91mobiles
91mobiles recently conducted a Smart TV Buyer Insights Survey in which we asked over 3,000 respondents about the TV they own, aspects they look at on a new TV, and their TV buying preferences.
Threats to mobile devices are more prevalent and increasing in scope and complexity. Users of mobile devices desire to take full advantage of the features
available on those devices, but many of the features provide convenience and capability but sacrifice security. This best practices guide outlines steps the users can take to better protect personal devices and information.
Unlocking Productivity: Leveraging the Potential of Copilot in Microsoft 365, a presentation by Christoforos Vlachos, Senior Solutions Manager – Modern Workplace, Uni Systems
Essentials of Automations: The Art of Triggers and Actions in FMESafe Software
In this second installment of our Essentials of Automations webinar series, we’ll explore the landscape of triggers and actions, guiding you through the nuances of authoring and adapting workspaces for seamless automations. Gain an understanding of the full spectrum of triggers and actions available in FME, empowering you to enhance your workspaces for efficient automation.
We’ll kick things off by showcasing the most commonly used event-based triggers, introducing you to various automation workflows like manual triggers, schedules, directory watchers, and more. Plus, see how these elements play out in real scenarios.
Whether you’re tweaking your current setup or building from the ground up, this session will arm you with the tools and insights needed to transform your FME usage into a powerhouse of productivity. Join us to discover effective strategies that simplify complex processes, enhancing your productivity and transforming your data management practices with FME. Let’s turn complexity into clarity and make your workspaces work wonders!
UiPath Test Automation using UiPath Test Suite series, part 4DianaGray10
Welcome to UiPath Test Automation using UiPath Test Suite series part 4. In this session, we will cover Test Manager overview along with SAP heatmap.
The UiPath Test Manager overview with SAP heatmap webinar offers a concise yet comprehensive exploration of the role of a Test Manager within SAP environments, coupled with the utilization of heatmaps for effective testing strategies.
Participants will gain insights into the responsibilities, challenges, and best practices associated with test management in SAP projects. Additionally, the webinar delves into the significance of heatmaps as a visual aid for identifying testing priorities, areas of risk, and resource allocation within SAP landscapes. Through this session, attendees can expect to enhance their understanding of test management principles while learning practical approaches to optimize testing processes in SAP environments using heatmap visualization techniques
What will you get from this session?
1. Insights into SAP testing best practices
2. Heatmap utilization for testing
3. Optimization of testing processes
4. Demo
Topics covered:
Execution from the test manager
Orchestrator execution result
Defect reporting
SAP heatmap example with demo
Speaker:
Deepak Rai, Automation Practice Lead, Boundaryless Group and UiPath MVP
GraphRAG is All You need? LLM & Knowledge GraphGuy Korland
Guy Korland, CEO and Co-founder of FalkorDB, will review two articles on the integration of language models with knowledge graphs.
1. Unifying Large Language Models and Knowledge Graphs: A Roadmap.
https://arxiv.org/abs/2306.08302
2. Microsoft Research's GraphRAG paper and a review paper on various uses of knowledge graphs:
https://www.microsoft.com/en-us/research/blog/graphrag-unlocking-llm-discovery-on-narrative-private-data/
Removing Uninteresting Bytes in Software FuzzingAftab Hussain
Imagine a world where software fuzzing, the process of mutating bytes in test seeds to uncover hidden and erroneous program behaviors, becomes faster and more effective. A lot depends on the initial seeds, which can significantly dictate the trajectory of a fuzzing campaign, particularly in terms of how long it takes to uncover interesting behaviour in your code. We introduce DIAR, a technique designed to speedup fuzzing campaigns by pinpointing and eliminating those uninteresting bytes in the seeds. Picture this: instead of wasting valuable resources on meaningless mutations in large, bloated seeds, DIAR removes the unnecessary bytes, streamlining the entire process.
In this work, we equipped AFL, a popular fuzzer, with DIAR and examined two critical Linux libraries -- Libxml's xmllint, a tool for parsing xml documents, and Binutil's readelf, an essential debugging and security analysis command-line tool used to display detailed information about ELF (Executable and Linkable Format). Our preliminary results show that AFL+DIAR does not only discover new paths more quickly but also achieves higher coverage overall. This work thus showcases how starting with lean and optimized seeds can lead to faster, more comprehensive fuzzing campaigns -- and DIAR helps you find such seeds.
- These are slides of the talk given at IEEE International Conference on Software Testing Verification and Validation Workshop, ICSTW 2022.
Encryption in Microsoft 365 - ExpertsLive Netherlands 2024Albert Hoitingh
In this session I delve into the encryption technology used in Microsoft 365 and Microsoft Purview. Including the concepts of Customer Key and Double Key Encryption.
GraphSummit Singapore | The Future of Agility: Supercharging Digital Transfor...Neo4j
Leonard Jayamohan, Partner & Generative AI Lead, Deloitte
This keynote will reveal how Deloitte leverages Neo4j’s graph power for groundbreaking digital twin solutions, achieving a staggering 100x performance boost. Discover the essential role knowledge graphs play in successful generative AI implementations. Plus, get an exclusive look at an innovative Neo4j + Generative AI solution Deloitte is developing in-house.
Dr. Sean Tan, Head of Data Science, Changi Airport Group
Discover how Changi Airport Group (CAG) leverages graph technologies and generative AI to revolutionize their search capabilities. This session delves into the unique search needs of CAG’s diverse passengers and customers, showcasing how graph data structures enhance the accuracy and relevance of AI-generated search results, mitigating the risk of “hallucinations” and improving the overall customer journey.
Observability Concepts EVERY Developer Should Know -- DeveloperWeek Europe.pdfPaige Cruz
Monitoring and observability aren’t traditionally found in software curriculums and many of us cobble this knowledge together from whatever vendor or ecosystem we were first introduced to and whatever is a part of your current company’s observability stack.
While the dev and ops silo continues to crumble….many organizations still relegate monitoring & observability as the purview of ops, infra and SRE teams. This is a mistake - achieving a highly observable system requires collaboration up and down the stack.
I, a former op, would like to extend an invitation to all application developers to join the observability party will share these foundational concepts to build on:
State of ICS and IoT Cyber Threat Landscape Report 2024 previewPrayukth K V
The IoT and OT threat landscape report has been prepared by the Threat Research Team at Sectrio using data from Sectrio, cyber threat intelligence farming facilities spread across over 85 cities around the world. In addition, Sectrio also runs AI-based advanced threat and payload engagement facilities that serve as sinks to attract and engage sophisticated threat actors, and newer malware including new variants and latent threats that are at an earlier stage of development.
The latest edition of the OT/ICS and IoT security Threat Landscape Report 2024 also covers:
State of global ICS asset and network exposure
Sectoral targets and attacks as well as the cost of ransom
Global APT activity, AI usage, actor and tactic profiles, and implications
Rise in volumes of AI-powered cyberattacks
Major cyber events in 2024
Malware and malicious payload trends
Cyberattack types and targets
Vulnerability exploit attempts on CVEs
Attacks on counties – USA
Expansion of bot farms – how, where, and why
In-depth analysis of the cyber threat landscape across North America, South America, Europe, APAC, and the Middle East
Why are attacks on smart factories rising?
Cyber risk predictions
Axis of attacks – Europe
Systemic attacks in the Middle East
Download the full report from here:
https://sectrio.com/resources/ot-threat-landscape-reports/sectrio-releases-ot-ics-and-iot-security-threat-landscape-report-2024/
State of ICS and IoT Cyber Threat Landscape Report 2024 preview
Web spam classification using supervised artificial neural network algorithms
1. Advanced Computational Intelligence: An International Journal (ACII), Vol.2, No.1, January 2015
21
WEB SPAM CLASSIFICATION USING SUPERVISED
ARTIFICIAL NEURAL NETWORK ALGORITHMS
Ashish Chandra, Mohammad Suaib, and Dr. Rizwan Beg
Department of Computer Science & Engineering, Integral University, Lucknow, India
ABSTRACT
Due to the rapid growth in technology employed by the spammers, there is a need of classifiers that are
more efficient, generic and highly adaptive. Neural Network based technologies have high ability of
adaption as well as generalization. As per our knowledge, very little work has been done in this field using
neural network. We present this paper to fill this gap. This paper evaluates performance of three supervised
learning algorithms of artificial neural network by creating classifiers for the complex problem of latest
web spam pattern classification. These algorithms are Conjugate Gradient algorithm, Resilient Back-
propagation learning, and Levenberg-Marquardt algorithm.
KEYWORDS
Web spam, artificial neural network, back-propagation algorithms, Conjugate Gradient, Resilient Back-
propagation, Levenberg-Marquardt, Web spam classification
1. INTRODUCTION
In this paper we are studying three artificial neural network (ANN) algorithms, which are
Conjugate Gradient learning algorithm, Resilient Back-propagation algorithm, and Levenberg-
Marquardt algorithm. All of these algorithms are supervised learning algorithms.
We are evaluating performance of these algorithms on the basis of classification results as well as
computational requirements in training. The application on which we are evaluating these
algorithms is web spam classification.
Web spam is one of the key challenges for search engine industry. Web spam are the web pages
that are created or manipulated to lead users from search engine result page to a target web page
and also manipulate search engine ranking of the target page.
2. RELATED WORK
Svore (2007) [1] devised a method for web spam detection based on content-based features and
the rank-time. They used SVM classifier with linear kernel.
Noi (2010) [2] proposed a combination of graph neural network and probability mapping graph
self organizing maps organized into a layered architecture. It was a mixture of unsupervised and
supervised learning approaches but the training time was computationally very expensive.
2. Advanced Computational Intelligence: An International Journal (ACII), Vol.2, No.1, January 2015
22
Erdelyi (2011) [3] achieved superior classification results in experiment using learning methods
LogitBoost and RandomForest with less computation hungry content features. They used 100,000
hosts from WebspamUK2007 and 190,000 hosts from DC2010 datasets and investigated the
trade-off between feature generation and spam classification accuracy. They proved that more
features improve performance but complex features such as PageRank improves the classification
accuracy marginally.
According to Biggio (2011) [4], SVM can be manipulated in adversarial classification tasks such
as spam filtering.
Similarly, Xiao (2012) [5] showed that injection of contaminated data in training dataset
significantly degrades the accuracy of the SVM.
This is well known that neural network perform better with noisy data. Similar is the case of
adversarial dataset of web spam.
3. ARTIFICIAL NEURAL NETWORK
An ANN is a collection of simple processing units which communicate with each other using a
large number of weighted connections. Each unit receives input from neighbour units or from
external source and computes output which propagates to other neighbours. There is also a
mechanism to adjust weights of the connections. Normally there are 3 types of units.
• Input Unit: which receives signal from external source.
• Output Unit: which sends data out of the network.
• Hidden Unit: which Receives and sends signals within the network.
Many of the units can work parallel in the system. ANN can be adapted to take a set of inputs and
produce a desired set of outputs. This process is known as learning or training. There are two
kinds of training in neural network.
• Supervised: The network is provided a set of inputs and corresponding output patterns,
called training dataset, to train the network.
• Unsupervised: The network trains itself by creating clusters of patterns. Here no prior set
training data is provided to the system.
3.1. Multi-Layer Perceptron (MLP)
MLP is a nonlinear feed forward network model which maps a set of inputs x into a set of outputs
y. It has 3 types of layers: input layer, output layer, and hidden layer.
Standard perceptron calculates a discontinuous function:
ݔሬሬሬԦ → ݂௦௧ (ݓ + (ݓሬሬԦ , ݔԦሻሻ (1)
smoothing is done using a logistic function to get
ݔሬሬሬԦ → ݂ (ݓ + (ݓሬሬԦ , ݔԦሻሻ (2)
where: ݂ (ݖሻ =
ଵ
ଵ ା ష
3. Advanced Computational Intelligence: An International Journal (ACII), Vol.
MLP is a finite acyclic graph where the nodes are neurons with logistic activation. Neurons of i
layer serves as input for neurons of (i+1)
network containing many neurons. All the
next layer.
Figure 1. A Multilayer Perceptron with one input, two hidden, and one output layer.
All connections are weighted with real number. Weight of connection i
output layer neurons have a bias weight w
Figure 1. A Neuron outputs an activation function applied over weighted sum of its all inputs
Each node of the network outputs an activation function applied over weighted sum of
inputs.
ܽ
The network output is given by the ai of the output neurons.
3.2. Neural Network Supervised Learning Algorithms
While implementing any learning algorithm, our objective is to reduce the Global
by:
Advanced Computational Intelligence: An International Journal (ACII), Vol.2, No.1, January 2015
MLP is a finite acyclic graph where the nodes are neurons with logistic activation. Neurons of i
layer serves as input for neurons of (i+1)th
layer. Very complex functions can be calculated with
network containing many neurons. All the neurons of one layer are connected to all neurons of
. A Multilayer Perceptron with one input, two hidden, and one output layer.
All connections are weighted with real number. Weight of connection i → j is wji. All hidden and
output layer neurons have a bias weight wi0.
. A Neuron outputs an activation function applied over weighted sum of its all inputs
Each node of the network outputs an activation function applied over weighted sum of
ܽ ൌ ݂( w୧ + w୧ ,୨
ୀଵ
X a୨ ሻ
The network output is given by the ai of the output neurons.
Neural Network Supervised Learning Algorithms
While implementing any learning algorithm, our objective is to reduce the Global Error E defined
ܧ ൌ
ଵ
E୮
ୀଵ
2, No.1, January 2015
23
MLP is a finite acyclic graph where the nodes are neurons with logistic activation. Neurons of ith
layer. Very complex functions can be calculated with
neurons of one layer are connected to all neurons of
. A Multilayer Perceptron with one input, two hidden, and one output layer.
. All hidden and
. A Neuron outputs an activation function applied over weighted sum of its all inputs
Each node of the network outputs an activation function applied over weighted sum of its all
(3)
Error E defined
(4)
4. Advanced Computational Intelligence: An International Journal (ACII), Vol.2, No.1, January 2015
24
Where P is the total size of training dataset, and Ep is the error for training pattern p. And
also,
ܧ ൌ
ଵ
ଶ
(O୧ − t୧ሻଶ
ୀଵ
(5)
where N is total number of output nodes, Oi is output of the ith
output node and ti is the target
output at the ith
output node. Every learning algorithm tries to reduce the global error by adjusting
weights and biases. Now first we will discuss the three algorithms one by one.
3.3. Conjugate Gradient (CG) Algorithm
It is basic back-propagation algorithm. It adjusts weights in the steepest descent direction i.e. the
most negative of the gradients. This is the direction in which the function is decreasing most
rapidly. It is observed that although the function decreases most rapidly along the negative of the
gradient direction, it does not always provide the fastest convergence. In the Conjugate Gradient
algorithm a search is done in conjugate directions, which generally provides the faster
convergence than the steepest descent direction[6].
Conjugate Gradient Algorithm adjusts step size in each iteration along the conjugate gradient
direction to minimize performance function. It searches the steepest descent direction on the first
iteration.
p0 = - g0 (6)
Then it performs line search to determine the optimal distance to move along the current search
direction by combining new steepest descent direction with the previous direction.
pk = - gk + βk pk - 1 (7)
where the constant βk is
β୩ ൌ
ౡ
ౡ
ౡషభ
ౡషభ
(8)
It is the ratio of the norm squared of current gradient to the norm squared of the previous
gradient[7].
3.4. Resilient Back-propagation (RB) Algorithm
Multilayer Perceptron networks typically use sigmoid transfer function in the hidden layer. It is
also known as squashing function because it compresses an infinite input range into finite output
range. The sigmoid function's slope approaches to zero as input gets large. It causes problem
when we use steepest descent to train network. Because since gradient may have very small value
so it may cause small changes in weights and biases even though the weight and biases are fairly
far away from the optimal values.
The purpose of the resilient back-propagation (RB) learning is to remove the harmful effect of the
magnitude of partial derivatives. It uses only the sign of the partial derivative to determine the
direction of the weight update. The magnitude of the weight change is calculated as following
rule [8]:
5. Advanced Computational Intelligence: An International Journal (ACII), Vol.2, No.1, January 2015
25
• Values of each weight and bias are increased when the derivative of performance
function with respect to that weight has same sign for two successful iterations.
• Values of each weight and bias are decreased when the derivative with respect to weight
changes sign from the previous iteration.
• If the derivative is zero the values of the weights and biases are remain same.
• When weights oscillate, values of weight change is reduced.
• If weights continuously change in the same direction, weight change is increased.
3.5. Levenberg-Marquardt (LM) Algorithm
It provides a numerical solution to the problem of minimizing a generally non-linear function,
over a space of parameters for the function. It is popular alternative to the Gauss-Newton method
of finding the minimum of a function. It approaches second order training speed without
computing the Hessian Matrix [9],
If the performance function is of the form of a sum of squares, then Hessian matrix can be
approximated as:
H = JT
J (9)
and the gradient is:
g = JT
e (10)
where J represents the Jacobian matrix containing the first derivatives of network error with
respect to weights and biases, and e is the vector of network errors.
The Jacobian matrix can be computed through a standard back-propagation technique that is
much complex than computing Hessian matrix. This approximated matrix is used in following
Newton like update
xk + 1 = xk - [ JT
J + µ I ]-1
JT
e (11)
When the scalar µ = 0, it becomes the Newton's method on approximated Hessian matrix. When µ
is large, it becomes gradient descent with small step size. The Newton's method is faster and more
accurate so algorithm shifts towards it. So the µ is decreased with each successful step (i.e. when
performance function reduces). It is increased when a tentative step would increase the
performance function. So the performance function always reduces at each iteration. It is more
powerful than conventional gradient descent technique.
LM Algorithm is very sensitive to initial network weights. It does not consider outliers in the data
which may lead to over-fitting noise. To avoid these situations, Regularization technique is used.
One of such technique is Bayesian Regularization.
3.6. Bayesian Regularization (BR)
It is used to overcome the problem of interpolating noisy data. Mackay (1992) proposed Bayesian
framework which can be directly applied to the NN learning problem. It allows to estimate
effective number of parameters actually used by the model i.e. the number of network weights
actually needed to solve a particular problem. Bayesian Regularization expands the cost function
to search not only for minimal error but for minimal error using minimal weights. By using
Bayesian Regularization one can avoid costly cross validation. It is particularly useful for the
6. Advanced Computational Intelligence: An International Journal (ACII), Vol.2, No.1, January 2015
26
problems that cannot, or would suffer, if a portion of available data were reserved to a validation
set. Furthermore, the regularization also reduces or eliminates the need for testing different
numbers of hidden neurons for a problem. Bayesian regularized ANN are difficult to be over
trained and over-fit.
4. CONFUSION MATRIX
Confusion matrix or contingency table is used to evaluate the performance of a machine learning
classifier. We used four attributes of confusion matrix while evaluating the performance of the
algorithms. These attributes are Sensitivity, Specificity, Efficiency and Accuracy. These attributes
are defined as following:
Sensitivity or True Positive Rate (TPR), also known as Recall Rate is given by:
Sensitivity ൌ
( ା ሻ
(12)
Specificity or True Negative Rate (TNR) is given by:
Speciϐicity ൌ
( ା ሻ
(13)
Efficiency is given by:
Efϐiciency ൌ
ୗୣ୬ୱ୧୲୧୴୧୲୷ାୗ୮ୣୡ୧ϐ୧ୡ୧୲୷
ଶ
ൌ
ୖାୖ
ଶ
(14)
Accuracy is given by:
Accuracy ൌ
ା
(ାሻ
(15)
where P is number of positive instances, N is number of negative instances, TP is number of
correctly classified positive instances, TN is number of correctly classified negative instances, FP
is number of incorrectly classified as positive instances and FN is number of incorrectly classified
as negative instances.
5. EXPERIMENT
We evaluated 3 supervised learning algorithm which uses back-propagation on multilayer
perceptron neural network. We checked the performance of these algorithm to automatically
detect web spam. These algorithms are Conjugate Gradient, Resilient Back-propagation and
Levenberg-Marquardt learning.
We created neural network with single hidden layer and used one neuron in the output layer. We
employed bipolar sigmoid function which has the output range of [ -1 , 1 ].
݂(ݔሻ ൌ
ଶ
ଵ ା షഀೣ − 1 (16)
We selected in the experiment:
Stopping Criteria as Number of Iterations θ = 100,
Learning Rate α = 0.1
7. Advanced Computational Intelligence: An International Journal (ACII), Vol.2, No.1, January 2015
27
Number of neurons in hidden layers = 10 (or 20 where specified).
We created a corpus of 368 instances of manually selected web pages, in which about 30%
instances were labelled as spam and rest of the pages were labelled as ham. To create training
dataset, we randomly selected about 80% of records from the corpus and for testing we used
remaining 20% of the records.
We extracted total 31 low cost quality features of the pages and categorized them in 3 categories:
URL (10 features), Content (16 features) and Link (5 features). We call these factors low cost
because they are computationally less expensive to be extracted. These features are listed in table
1.
Table I. Low Cost Quality Features of a Web Page
URL Features
SSL Certificate
Length of URL
URL is not a sub-domain
TLD is authoritative (*.gov, *.edu, *.ac)
Domain contains more than 2 same consecutive
alphabet)
Sub-domain is more than level 3 (e.g.
mocrosoft.com.mydomain.net)
Domain contains many digits and special symbols
IP address instead of Domain Name
Alexa Top 500 site.
Domain Length
Content Features
HTML Length
Word Count
Text Character Length
Number of Images
Description Length
Existence of H2
Existence of H1
Video Integration
Number of Ads
Title Character Length
Compression Ratio of Text
Ratio of text to HTML
Presence of alt text for images
Presence of obfuscated JavaScript code (pop-ups,
redirections etc)
% of Call to Action in the Text
Percentage of Stop words in Text
Links Features
Number of Internal links
Internal link is self referential
Number of External links
Fraction of anchor text / Total Text
Word Count in Anchor Text
8. Advanced Computational Intelligence: An International Journal (ACII), Vol.2, No.1, January 2015
28
To test the performance of each algorithm, we trained, tested and obtained 10 performance result
values for each category and calculated the average.
We used Accord.net and Aforge.net libraries to create neural network.
6. RESULT
We have created tables (Table II to Table VIII) to show the performance results of each
algorithm. Consider the number of neurons in hidden layer as 10 unless specified. The values in
the tables represent average of 10 experimental readings of each category. The values in
underlined show the best results.
Table 2. Using Url Features Only (10 Features)
Algorithm Sensitivity
(TPR)
Specificity
(TNR)
Efficiency Accuracy Training Time
(seconds)
CG 0.9153 0.3088 0.6121 0.5716 0.149
RB 0.4653 0.9117 0.6885 0.7183 0.168
LM 0.6884 0.7647 0.7265 0.7316 2.723
LM+BR 0.4224 0.9264 0.6744 0.7090 5.790
Table 3. Using Content features Only (16 Features)
Algorithm Sensitivity
(TPR)
Specificity
(TNR)
Efficiency Accuracy Training Time
(seconds)
CG 0.7384 0.8823 0.8104 0.8200 0.198
RB 0.7269 0.9205 0.8237 0.8366 0.191
LM 0.6615 0.8441 0.7528 0.7650 9.685
LM+BR 0.6116 0.9380 0.7748 0.7952 16.445
Table 4. Using Links features Only (5 Features)
Algorithm Sensitivity
(TPR)
Specificity
(TNR)
Efficiency Accuracy Training Time
(seconds)
CG 0.9230 0.6382 0.7806 0.7616 0.113
RB 0.6692 0.9117 0.7904 0.8066 0.157
LM 0.7230 0.9088 0.8159 0.8282 1.923
LM+BR 0.7038 0.8880 0.7959 0.8080 2.722
Table 5. Using URL + Links Features (15 Features)
Algorithm Sensitivity
(TPR)
Specificity
(TNR)
Efficiency Accuracy Training Time
(seconds)
CG 0.9538 0.7970 0.8754 0.8650 0.191
RB 0.8076 0.9823 0.8950 0.9066 0.187
LM 0.8115 0.9352 0.8734 0.8816 3.744
LM+BR 0.9115 0.9558 0.9337 0.9366 13.458
9. Advanced Computational Intelligence: An International Journal (ACII), Vol.2, No.1, January 2015
29
Table 6. Using URL + Content features (26 Features)
Algorithm Sensitivity
(TPR)
Specificity
(TNR)
Efficiency Accuracy Training Time
(seconds)
No. of neurons in hidden layer=10
CG 0.8846 0.9264 0.9055 0.9083 0.266
RB 0.8538 0.9764 0.9151 0.9233 0.190
LM 0.8807 0.8823 0.8815 0.8816 10.743
LM+BR 0.7115 0.9911 0.8513 0.8700 59.957
No. of neurons in hidden layer=20
CG 0.8730 0.9558 0.9144 0.9200 0.539
RB 0.8576 0.9823 0.9200 0.9283 0.312
LM 0.8884 0.9205 0.9045 0.9066 32.789
LM+BR 0.7692 1.000 0.8846 0.9000 304.134
Table 7. Using Content + Links features (21 Features)
Algorithm Sensitivity
(TPR)
Specificity
(TNR)
Efficiency Accuracy Training Time
(seconds)
CG 0.7269 0.8705 0.7987 0.8083 0.230
RB 0.7593 0.9088 0.8340 0.8438 0.188
LM 0.7038 0.8441 0.7739 0.7833 7.940
LM+BR 0.6846 0.9470 0.8158 0.8333 36.674
Table 8. Using URL + Content + Links features (31 Features)
Algorithm Sensitivity
(TPR)
Specificity
(TNR)
Efficiency Accuracy Training Time
(seconds)
No. of neurons in hidden layer=10
CG 0.8500 0.9676 0.9088 0.9166 0.297
RB 0.8769 0.9735 0.9252 0.9316 0.214
LM 0.8653 0.9029 0.8841 0.8866 13.146
LM+BR 0.6923 0.9852 0.8388 0.8583 92.248
No. of neurons in hidden layer=20
CG 0.8384 0.9823 0.9104 0.9200 0.575
RB 0.8807 0.9588 0.9197 0.9250 0.377
LM 0.8653 0.9264 0.8959 0.9000 45.499
LM+BR 0.8115 0.9882 0.8998 0.9116 357.580
Data from the tables suggest that Conjugate Gradient (CG) Algorithm gave best TPR (Sensitivity)
in most of the categories, whereas Levenberg-Marquardt algorithm with Bayesian Regularization
(LM+BR) gave best TNR (Specificity) in most of the categories.
The overall best classification performance was achieved in most of the categories by Resilient
Back-propagation (RB) algorithm in both Efficiency and Accuracy measures.
The training time was found lowest in most of the cases for Resilient Back-propagation (RB) so
we can say that it is not only best in classification, it is fastest algorithm as well. Conjugate
10. Advanced Computational Intelligence: An International Journal (ACII), Vol.2, No.1, January 2015
30
Gradient (CG) works fast when number of inputs are less, but when number of inputs are
increased, its becomes slower than RB. The slowest algorithm observed was Levenberg-
Marquardt with Bayesian Regularization (LM+BR) in all of the cases in the experiment.
The overall classification performance of each algorithm improved with increased number of
page quality factors but training time also increased. The cases where the number of quality
factors were high, increasing number of neurons improved the classification performance of
algorithms but it also increased the training time.
7. CONCLUSION
In the experiment we can conclude that the Resilient Back-propagation algorithm is fastest and
performs best in both Efficiency and Accuracy measures. Conjugate Gradient algorithm gives
best sensitivity, whereas Levenberg-Marquardt algorithm with Bayesian Regularization gives best
specificity but it is the slowest when training time is considered.
Classification performance of each algorithm improves with increased input factors. If the
number of factors are high, increased number of neurons improves the performance of algorithm.
The training time increases for all algorithms when either number of inputs are increased or
number of neurons in hidden layer are increased.
REFERENCES
[1] Svore, K.M., Wu, Q., Burges, C.J.: "Improving web spam classification using rank-time features," in
Proc. of the 3rd AIRWeb, Banff, Alberta, Canada (2007) 9–16.
[2] Noi, L.D., Hagenbuchner, M., Scarselli, F., Tsoi, A., "Web spam detection by probability mapping
graphsoms and graph neural networks," in Proc. of the 20th ICANN, Thessaloniki, Greece (2010)
372–381.
[3] M. Erdelyi, A. Garzo, and A. A. Benczur, "Web spam classification: a few features worth more," in
Proceedings of the 2011 Joint WICOW/AIRWeb Workshop on Web Quality, WebQuality'11,
Hyderabad, India, 2011.
[4] B. Biggio, B. Nelson, and P. Laskov, "Support vector machines under adversarial label noise," in
JMLR: Workshop and Conference Proceedings 20, Taoyuan, Taiwan, 2011, pp. 97–112.
[5] H. Xiao, H. Xiao, and C. Eckert, "Adversarial label flips attack on support vector machines,"
presented at the 20th European Conference on Artificial Intelligence (ECAI), Montpellier, France,
2012.
[6] Adeli H & Hong SL, "Machine learning neural networks genetic algorithms and fuzzy systems" (John
Wiley & Sons Inc., New York, NY, USA) 1995.
[7] Fletcher R & Reeves CM, Computer J, 7 (1964) 149-153.
[8] Reidmiller M & Brain H, "A direct adaptive method for faster back-propagation learning: The
RPROP algorithm," Proc IEEE Int. Conf. Neural Networks, 1993.
[9] More JJ, in "Numerical Analysis", edited by Watson GA, Lecture Notes in Mathematics 630,
(Springer Verlog, Germany) 1997, 105-116.